Frequently Asked Questions

Emergence Field Labs (EFL) is an evaluation and learning infrastructure organization built for nonprofits, grassroots organizations, and social sector practitioners. We build AI-powered tools that make rigorous, community-controlled evaluation accessible to any organization, not just those with research teams or big evaluation budgets.

The premise is simple: social systems fail not because of bad people or bad intentions, but because of weak feedback loops. Organizations collect data constantly but rarely turn it into learning. EFL exists to fix that, to build the infrastructure that helps communities structure their own experience, own their own data, and participate in a network of collective knowledge.

EFL is built for organizations doing the work: program managers, practitioners, field staff, and small evaluation teams inside nonprofits, foundations, civic organizations, community development groups, and academics. Specifically:

• Organizations and individuals that need to measure their impact but lack a full-time research team
• Programs that serve multiple funders and need to attribute data across grants without duplicating work
• Grassroots and community-based organizations that want to own and control their evaluation data
• Networks and hubs that want to aggregate learning across member organizations while preserving local data sovereignty
• International development and peacebuilding organizations working in complex, multi-stakeholder environments

You don't need to be an evaluator or a data scientist. EFL is designed for practitioners, not methodologists.

The social sector has a measurement problem, but it's not the one most people think. The sector doesn't lack data. It lacks infrastructure that actually turns data into learning. Most organizations are stuck in a loop: collect data to justify grants already written, produce reports that don't inform decisions, and watch institutional knowledge walk out the door when staff leave.

At the same time, the tools available (expensive survey platforms, siloed databases, generic analytics software) were built for corporate research teams, not community organizations. They're hard to use, they extract value, and they own your data.

EFL was founded by practitioners with sixteen years of experience in international development and monitoring & evaluation, including field research across 30+ countries and coordinating the evacuation of 1,200 vulnerable Afghans after Kabul fell in 2021. We built EFL because we lived the problem. We know what happens when communities generate data that serves funder narratives instead of community learning.

Most evaluation tools solve a data collection problem. EFL solves a learning infrastructure problem. That means:

• Your data is yours — permanently. We don't sell it, use it to track you, or share it for advertising. It belongs to your organization from the moment you create it.
• AI assists, it doesn't replace. MERLin, our AI research assistant, is designed to build your organization's evaluative capacity, not substitute for it. Every feature is built to keep human judgment at the center.
• Qualitative and quantitative are treated as co-equal evidence. Field observations, voice notes, and open-ended responses aren't footnotes. They're first-class intelligence.
• The platform is built for the field, not the boardroom with voice-first collection, low-friction design, and mobile-accessible for populations with limited text access.
• Network learning without extraction. Organizations can learn from each other's work without surrendering their data.

We're in beta. We're honest about what's built vs. what's on the roadmap. If you want the plain-language answer on current vs. planned features, just ask.

MERLin is EFL's AI research assistant, embedded across all five platform stages: Diagnostic, Framework Builder, Research Approach, Survey Design, and Results & Analysis. It's not a chatbot. It's a structured thinking partner that guides you through the evaluation lifecycle.

MERLin is built on Anthropic's Claude models and governed by a detailed prompt specification that controls its behavior across every section of the platform.

MERLin is designed around three core behavioral commitments:

• Human-first. Users always write or respond before MERLin engages. You think first. MERLin responds to what you've said, not what it assumes you mean.
• Socratic, not prescriptive. MERLin asks questions to surface your thinking. It doesn't hand you frameworks; it helps you build them. That deliberate friction is the product.
• Intentional friction at high-stakes decisions. When you're making consequential choices about your evaluation framework or assumptions, MERLin slows the process down. Speed is not the goal. Learning is.

This is rooted in a core design principle: the difference between agency and autonomy. Agency is the capacity to achieve outcomes. Autonomy is the right to set the criteria by which outcomes are judged. EFL is designed to build your organization's agency, its evaluative capacity, not to substitute for your judgment.

It can help you build it. It won't build it for you, and that's by design.

One of the central risks in AI-assisted evaluation is what we call constitutional drift: when AI efficiency gradually displaces your organization's own standards, values, and judgment. An evaluation framework that MERLin produced wholesale would reflect what MERLin thinks your program should look like, not what your community actually believes about how change happens.

That's not evaluation. That's performance. MERLin is designed to make you a better evaluator, not to make evaluators unnecessary.

Theory of Change is the first fully built framework in EFL's platform. MERLin's eventual diagnostic architecture will support 30+ frameworks across six framework families:

• Causal Logic — Theory of Change, Logframe, Logic Model, Results Framework
• Participatory & Community-Centered — Most Significant Change, Outcome Harvesting, Appreciative Inquiry, ABCD
• Complexity-Aware & Adaptive — Developmental Evaluation, Realist Evaluation, Contribution Analysis, Outcome Mapping
• Systems-Level — Systems Change Evaluation, Collective Impact, Causal Loop Diagramming
• Qualitative & Interpretive — Case Study, Narrative Inquiry, Phenomenological Inquiry
• Experimental — RCT, Quasi-Experimental, Pre-Post / Longitudinal Cohort

MERLin will diagnose which framework best fits your situation, based on your program type, funder requirements, organizational capacity, and the nature of the change you're working toward.

No. Never. This is not a policy preference that could change with a new business model or a new leadership team. It is a structural commitment encoded into EFL's governance architecture.

EFL will never sell, license, or transfer your organization's data to any of the following, under any circumstances:

• Government agencies, local, state, federal, or foreign
• Law enforcement or intelligence agencies
• Corporations, advertisers, or marketing firms
• Academic researchers or institutions
• Consulting firms, think tanks, or policy organizations
• Funders, including funders of your own programs
• Other nonprofits or civil society organizations
• Any third party not explicitly authorized in writing by your organization

The EFL Data Trust — the nonprofit entity that holds your data — has legal authority over how data is handled. The operating company builds the tools, but cannot access or commercialize community data outside of what the Data Trust permits. This isn't a policy preference that could quietly change; it's a structural constraint built into how EFL is organized.

No. Your organization's data will never be used to train any AI model that EFL uses.

EFL accesses AI providers exclusively through commercial APIs under paid plans. We only work with providers whose commercial terms explicitly prohibit using customer data for model training. That commitment applies to every provider we use — now and in the future.

If a provider ever changes their policy, we will either renegotiate under a Zero Data Retention agreement, switch providers, or remove that integration. We will notify affected organizations before any such change takes effect.

Access is governed by role-based permissions and full audit logging. Within EFL:

• Your organization's administrators control who on your team can access your data and at what permission level
• EFL engineers have technical access to the underlying database in the current Firebase-centralized architecture; we do not hide this fact
• All EFL staff access to any client data is logged with a full audit trail, and access is governed by the EFL Data Trust's policies, not by the operating company's discretion
• No EFL employee may access client data for commercial, analytical, or training purposes — only for technical support explicitly requested by the client organization

As EFL moves toward a federated architecture, even technical staff access will be architecturally constrained: queries will be submitted to data stores without raw data ever moving to a central server. We'll tell you when that architecture is live.

EFL collects only what's necessary to run the platform. We don't build profiles on your respondents, track behavior for advertising, or pull in data from outside sources.

What we collect:

• Organizational profile data: name, sector, program descriptions, geographic context
• Evaluation framework content: Theory of Change, indicators, assumptions
• Survey instruments and response data: structured and qualitative responses
• Field notes: voice recordings and text notes captured by program staff
• User account data: email addresses and role assignments

What we do not collect:

• No personally identifiable information about survey respondents beyond what your organization explicitly chooses to capture
• No device tracking, behavioral profiling, or advertising data of any kind
• No data from third-party sources about your communities or beneficiaries

EFL currently runs on Firebase (Google Cloud Infrastructure), hosted in the United States. Organizations with specific data residency needs should contact us to discuss options.

Many EFL users work in contexts where data exposure carries real risk: democracy and civic programs, peacebuilding work, programs serving undocumented populations, whistleblower support, and post-conflict community development. We design for those users, not against them.

• Surveys go to known respondent lists with unique identifiers. No open survey links that expose respondent identity through participation alone.
• We build friction into data collection design that encourages organizations to collect only what they need for learning, not for reporting performance.
• Our long-term federated architecture distributes data sovereignty to each organization. A federated system is a security system; you can't extract what isn't centralized.
• We design toward minimizing what we hold rather than maximizing what we can collect. The less we have, the less can be breached, subpoenaed, or misused.

If your program works with populations at elevated risk and you want to discuss your specific threat model before using EFL, contact us. These conversations make us better.